Skip to content

19th November 2020

Jump to bottom
agrimwood edited this page Nov 19, 2020 · 5 revisions

Meeting 19/110/2020

Attendance:

Alex Grimwood AG
Zac Baum ZB
Adria Casamitjana AC
Yunguan Fu YF
Yipeng Hu YH
Zhe Min ZM
Nina Montana Brown NB
Remi Delauny RD
Mark Pinnock MP
Shaheer Saeed SS

Apologies:

Stefano Blumberg SB
Ester Bonmati EB
Matt Clarkson MC
Qianye Yang QY

Agenda:

YH: Empty ticket review
YH: How to make DeepReg work for specific applications (demo review) ZB: Release scheduling

Ticket Review:

  • 459: assigned to MP
  • 468: YF
  • 474: NB need to consistently implement SS's visualisation tools. Assigned to SS
  • 476: YH requires further discussion
  • 515: assigned to NB to fix link, close issue and open a new issue for any follow on conversations
  • 526: assigned to NB. YH to discuss with AC and SB also.
  • 535: assigned to YF
  • 536: assigned to YH and AG

Improving Demos:

  • YH: urges people to explore hyperparameter tuning, adding data, enlarging networks etc. to improve performance
  • validation against other techniques is also favourable
  • ties into issue #536, but requires its own ticket.
  • AG to discuss offline with YH, and to raise ticket. Must identify a demo and to improve performance in first instance.

Release Scheduling:

  • Periodic Github release scheduling hasn't been tested yet.
  • How to define schedule?
  • YF: new release after minor typos
  • NB: agreed
  • Consensus is to agree new releases as part of regular dev meetings
  • Add release procedure to docs once the process has been tested

Other business:

  • AC: implementing spline, DDF, augmentation. Requires new input parameters, including possibility to change config file. This is not backwards compatible. What is the best way to proceed?
    • YH prefers backwards compatibility. NB and YF to discuss. Requires discussion with AC
  • AC & NB: there's scope for presenting classical registration functionality/wrappings to the user. Ticket to be raised
  • AC: scope for a validation exercise/publication comparing DeepReg against ClassicReg across a range of applications
    • YH: neuro and longitudinal data are most appealing research targets.
    • AC: willing to develop the idea further
  • NB: Advent of code. Suggests a sprint towards end of term instead of the bi-weekly meeting.
    • NB compiling a list of possible tasks/topics.

tables: for google docs --> markdown, use tablesgenerator.org
formatting: for google docs --> markdown, use stackedit

Clone this wiki locally